CN110162606B

CN110162606B - Session agent learning model service selection for resolving client service requests

Info

Publication number: CN110162606B
Application number: CN201910111869.6A
Authority: CN
Inventors: G·西雅诺; L·皮切特
Original assignee: International Business Machines Corp
Current assignee: International Business Machines Corp
Priority date: 2018-02-13
Filing date: 2019-02-12
Publication date: 2023-05-09
Anticipated expiration: 2039-02-12
Also published as: CN110162606A; US20190253517A1; US10834227B2

Abstract

The present disclosure relates to session proxy learning model service selection for resolving client service requests. Approaches are described with respect to session proxy learning model service selection in response to client service requests. According to this approach, the model advisor can predict which model service(s) can most properly address client problems. Methods related to this approach may include: a client service request associated with a client system and data associated with the client service request are received, wherein the data associated with the client service request includes at least one keyword extracted from the client service request and the context information. The method may further comprise: at least one model service is selected from a plurality of model services associated with the session proxy learning model based on data associated with the client service request. The method may further comprise: in response to the client service request, a selection of at least one model service is provided to the client system.

Description

Session agent learning model service selection for resolving client service requests

Technical Field

Various embodiments described herein relate generally to a session proxy learning model (e.g., a robot (bot) model, a chat robot (chatbot) model, or a session model). More specifically, various embodiments describe session proxy learning model service selection in response to client service requests.

Background

Cognitive computing technology is increasingly often used to solve client problems. The computer-based session model may provide one or more services to facilitate resolution of such client-side problems. Determining the best option(s) to address a given client service request conventionally requires querying multiple providers, given the diversity of available model services. Thus, conventional model service selection may require significant computational and/or economic costs.

Disclosure of Invention

Various embodiments described herein provide techniques for selecting session proxy learning model service(s) in response to a client service request. An associated method may include: a client service request associated with a client system and data associated with the client service request are received. The data associated with the client service request may include at least one keyword extracted from the client service request and the context information. The method may further comprise: at least one model service is selected from a plurality of model services associated with the session proxy learning model based on data associated with the client service request. The method may further comprise: a selection of at least one model service is provided to the client system in response to the client service request. In an embodiment, the context information may include data associated with the client system and information regarding at least one service request related to the client service request.

According to an embodiment, the step of selecting at least one model service among the plurality of model services may comprise: in response to determining that a stored service request of the plurality of stored service requests referenced in the predictive database corresponds to the client service request, at least one model service associated with the stored service request is selected from the plurality of model services. The step of selecting at least one model service among the plurality of model services may further comprise: in response to determining that none of the plurality of stored service requests corresponds to the client service request, at least one model service is selected from the plurality of model services by evaluating the client service request based on a plurality of predefined factors. In an embodiment, the step of determining that a stored service request of the plurality of stored service requests corresponds to a client service request may comprise: the stored service request is determined to meet input criteria associated with the client service request, and the stored service request is determined to meet at least one predetermined similarity threshold with respect to the client service request.

According to an embodiment, the step of selecting at least one model service among the plurality of model services by evaluating the client service request based on a plurality of predefined factors may comprise: the method includes classifying data associated with the client service request and measuring the classified data based on social media trends. Optionally, the step of selecting at least one model service among the plurality of model services by evaluating the client service request based on a plurality of predefined factors may further comprise: the categorized data is measured based on at least one cost variable and/or the categorized data is measured based on the client configuration. The step of selecting at least one model service among the plurality of model services by evaluating the client service request based on a plurality of predefined factors may further comprise: at least one model service is selected from the plurality of model services in response to measuring the classified data, and an association between the client service request and the selected at least one model service is stored in the predictive database. In an embodiment, the step of selecting at least one model service among the plurality of model services in response to measuring the classified data may comprise: model service scores calculated for the classified data, respectively, in the context of each of the plurality of model services are compared.

In an embodiment, the step of measuring the categorized data based on social media trends may include: an emotional score for at least one aspect of the categorized data is calculated within the set of social media providers in the context of each of the plurality of model services. In yet another embodiment, the step of measuring the categorized data based on social media trends may include: popularity scores for at least one aspect of the categorized data are calculated within the social media provider collection in the context of each of a plurality of model services. In yet another embodiment, the step of measuring the categorized data based on social media trends may include: a scope score for at least one aspect of the categorized data is calculated within the social media provider collection in the context of each of the plurality of model services.

Additional embodiments include a computer program product comprising a computer-readable storage medium having program instructions embodied therewith. According to such embodiments, the program instructions may be executable by a computing device to cause the computing device to perform one or more steps of the methods recited above. Yet another embodiment includes a system having a processor and a memory storing an application program that, when executed on the processor, performs one or more steps of the method recited above.

Drawings

So that the manner in which the above recited aspects are attained and can be understood in detail, a more particular description of the embodiments, briefly summarized above, may be had by reference to the appended drawings.

It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.

FIG. 1A illustrates a computing infrastructure depicting a client-server configuration relative to a session proxy learning model, in accordance with an embodiment.

FIG. 1B illustrates a schematic diagram of learning a model with respect to a session proxy, according to an embodiment.

Fig. 2 illustrates a method of resolving a client service request via a session proxy learning model according to an embodiment.

Fig. 3 illustrates a method of selecting at least one model service among a plurality of session proxy learning model services according to an embodiment.

Fig. 4 illustrates a method of determining whether a stored service request of a plurality of stored service requests corresponds to a client service request, according to an embodiment.

Fig. 5 illustrates a method of selecting at least one model service among a plurality of session proxy learning model services by evaluating a client service request based on a plurality of predefined factors, according to an embodiment.

FIG. 6 illustrates a method of measuring categorized data associated with client service requests based on social media trends according to an embodiment.

Fig. 7 illustrates a schematic diagram of calculating model service scores for classified data associated with a client service request in the context of each of a plurality of session proxy learning model services, according to an embodiment.

Detailed Description

Various embodiments described herein relate to techniques related to session proxy learning model service selection in response to client service requests. According to various embodiments, at least one model service may be selected in response to a client service request. Such model service selection may constitute a prediction of which model service(s) may most appropriately address the client service request. The selected model service(s) may be provided directly by the model server system or alternatively may be accessed via one or more external service providers.

The various embodiments described herein may be advantageous over conventional techniques. In particular, by facilitating selection of at least one model service in response to a client service request, various embodiments enable prediction of at least one model service, thus eliminating the need to query multiple model service providers. Accordingly, predictions in accordance with various embodiments may minimize any time costs, computational costs, and/or financial costs required to individually query or otherwise occupy multiple payment service providers. Furthermore, the selection of at least one model service according to various embodiments may enable prediction of at least one model service available directly via the model server system and/or at least one model service that may be provided in the form of software as a service outside the model server system. Furthermore, through the use of various evaluation techniques, including predefined factors such as social media trends, various embodiments may improve accuracy in selecting at least one model service over conventional techniques, thus potentially reducing time and computational overhead. Some of the various embodiments may not include all of these advantages, and not all of the embodiments may require them.

Hereinafter, various embodiments of the present invention are referred to. However, it should be understood that the invention is not limited to specific described embodiments. Rather, any combination of the following features and elements, whether related to different embodiments or not, is contemplated to implement and practice the invention. Furthermore, although various embodiments may achieve advantages with respect to other possible solutions and/or with respect to the prior art, whether a particular advantage is achieved by a given embodiment is not limiting. Thus, the following aspects, features, embodiments and advantages are merely illustrative and are not considered elements or limitations of the appended claims except where explicitly recited in a claim(s). Likewise, references to "the invention" should not be construed as an generalization of any inventive subject matter disclosed herein and should not be considered to be an element or limitation of the appended claims except where explicitly recited in a claim(s).

The present invention may be any possible system, method, and/or computer program product integrating a level of technical detail. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to perform various aspects of the invention.

The computer readable storage medium may be a tangible device that can hold and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: portable computer disks, hard disks, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static Random Access Memory (SRAM), portable compact disk read-only memory (CD-ROM), digital Versatile Disks (DVD), memory sticks, floppy disks, mechanical coding devices, punch cards or in-groove structures such as punch cards or grooves having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media, as used herein, are not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., optical pulses through fiber optic cables), or electrical signals transmitted through wires.

The computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device or to an external computer or external storage device over a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers and/or edge servers. The network interface card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium in the respective computing/processing device.

Computer readable program instructions for performing the operations of the present invention may be assembly instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, c++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present invention are implemented by personalizing electronic circuitry, such as programmable logic circuitry, field Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), with state information for computer readable program instructions, which can execute the computer readable program instructions.

Various aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable medium having the instructions stored therein includes an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Various embodiments described herein may be provided to end users through a cloud computing infrastructure. Cloud computing generally refers to providing scalable computing resources, i.e., services, over a network. More formally, cloud computing may be defined as providing abstract computing power between computing resources and their underlying technology architecture (e.g., servers, storage, networks), thereby enabling convenient on-demand network access to a shared pool of configurable computing resources that may be quickly provided and released with minimal management effort or service provider interaction. Thus, cloud computing allows users to access virtual computing resources (e.g., storage, data, applications, and even completed virtualized computing systems) in the cloud without regard to the underlying physical systems (or locations of those systems) used to provide the computing resources.

Typically, cloud computing resources are provided to users on a pay-per-use basis, wherein users pay only for actual use of computing resources (e.g., the amount of storage space consumed by the user or the number of virtualized systems instantiated by the user). A user can access any resource present in the cloud at any time and from anywhere over the internet. In the context of the present disclosure, a workload of a client computing system or server system running an application according to various embodiments described herein may be deployed to a computing cloud. Further, cloud-based database systems, virtual machines, and various other server applications may be used to manage these workloads.

Particular embodiments describe techniques for selecting session proxy learning model service(s) for resolving client service requests. However, it is to be understood that the techniques described herein may be adapted for various purposes other than those specifically described herein. Accordingly, references to specific embodiments are included for purposes of illustration and not limitation.

FIG. 1A illustrates a computing infrastructure 100 according to an embodiment. As shown, the computing infrastructure 100 includes a client computing system 105, a model system interface 135, a model server system 137, external sources 165 (e.g., social networks and/or social applications), and external service providers 167 (e.g., providers of external chat bot services or other external model-based services in the form of software-as-a-service), which may each be connected to a communication network 175. Although client computing system 105 and model server system 137 are illustrated in fig. 1A, computing infrastructure 100 may include any number of computing systems and more particularly may include any number of client systems and/or server systems.

Illustratively, the client computing system 105 may include a memory 107, a storage 109, a client input/output (I/O) device interface 111, a Central Processing Unit (CPU) 113, and a client network interface 115, all of which may be interconnected via an interconnect 117 (e.g., a bus). Although shown as a single computing system, client computing system 105 is included to represent a single client or multiple clients. In an embodiment, the client computing system 105 may be a thin client. Memory 107 may include a client application 119. Client application 119 may be an online or web application configured to interface with model server system 137 and other computing systems. Client application 119 may include application interface 121. In particular, the application interface 121 may include a model instance 123 (e.g., a conversation agent, a bot instance, or a chat bot instance) for interfacing with a conversation agent learning model via a model server system 137. Model instance 123 may include a Graphical User Interface (GUI) having a plurality of elements, e.g., text boxes, radio buttons, drop-down menus, etc., for facilitating interactions between the client and the session proxy learning model. Additionally or alternatively, model instance 123 may include a command line interface for facilitating interactions between clients and session proxy learning models. The storage 109 may include client application data 125 associated with a client application 119. The client application data 125 may include information regarding any interactions between the client and the session proxy learning model via the model instance 123. The client I/O device interface 111 may be communicatively coupled to one or more client I/O devices 127. The CPU 113 is included in fig. 1A to represent a single CPU, a plurality of CPUs, a single CPU having a plurality of processing cores, and the like. Client network interface 115 may receive data from model server system 137 via network 175 and may transmit the data to model server system 137. The client computing system 105 may be configured to interface with other server system(s) within the computing infrastructure 100 and/or may be configured to interface with server system(s) external to the computing infrastructure 100.

Model system interface 135 may serve as a service manager with respect to client computing system 105, as well as any other computing system that interfaces with model server system 137. Model system interface 135 may facilitate communication to and from model server system 137 via network 175. In particular, model system interface 135 may receive data from client computing system 105 via network 175 and may transmit data to client computing system 105. According to one embodiment, model system interface 135 may be a separate computing system from model server system 137, including a separate processor, memory, and/or storage device. Alternatively, the model system interface 135 may be a component (e.g., a software application and/or hardware entity) of the model server system 137.

Model server system 137 may include memory 139, storage 141, server I/O device interface 143, and CPU 145, all of which may be interconnected via interconnect 147 (e.g., a bus). Although shown as a single computing system, model server system 137 is included to represent a single server system or multiple server systems. Model server system 137 may be an automated service management system. Memory 139 may include a model advisor 149 and a database management system (DBMS) 151. As further described herein, model advisor 149 may be an application capable of selecting or facilitating selection of a session proxy learning model service in response to a client service request in accordance with various embodiments described herein. The DBMS151 is included to represent a single database management system or multiple database management systems. Storage 141 may include model advisor data 153, model policies 155, a repository collection 157, and a model prediction database 159. Model advisor 149 can generate and process model advisor data 153 based on interactions with client computing system 105 and/or other computing system(s). Model advisor data 153 can include aspects based on a corpus of information. The corpus of information may include records from a repository set (e.g., repository set 157) and external data, including data from a social network or application (e.g., from external sources 165). The model advisor data 153 further can include a plurality of model services, which can include an "on-preset" model service. Model advisor data 153 can further include one or more references to external model services provided by external service provider 167. Model advisor data 153 can further include information regarding such external model services. Such external model services may be in the form of software as a service. The model advisor 149 may send database query requests (e.g., client service requests or other queries received from the client computing system 105) to the DBMS151, and the DBMS151 may process the queries, for example, based on data in the repository set 157 and/or based on data in the model prediction database 159. The DBMS151 may include a software application configured to manage a repository set 157 and/or a model prediction database 159.

The repository collection 157 may include one or more relational databases. Further, the repository collection 157 may include one or more ontology trees or other ontology structures. Although fig. 1A illustrates four repositories in the repository set 157, the computing infrastructure 100 may include any number of repositories. Model prediction database 159 may be a relational database. Alternatively, the model prediction database 159 may include one or more ontology trees or other ontology structures. According to an embodiment, DBMS151 may send one or more requests to a remote database (not shown) via network 175. The DBMS151 further may include a software application configured to manage the model prediction database 159. Model prediction database 159 may include associations between stored client service requests and corresponding model services. The server I/O device interface 143 can be communicatively coupled to one or more server I/O devices 161.CPU 145 is included in FIG. 1A to represent a single CPU, multiple CPUs, a single CPU having multiple processing cores, and the like. Model advisor 149 can accept requests sent by client computing system 105 to model server system 137 via model system interface 135. Model advisor 149 may further transmit data to client computing system 105 via model system interface 135. External sources 165 and/or external service providers 167 may transmit external data to model advisor 149 via model system interface 135 or may otherwise communicate with model advisor 149.

In an embodiment, one or more functions related to selecting a session proxy learning model service in response to a client service request may be performed external to model server system 137. According to this embodiment, model advisor 149 may communicate with application(s) external to model server system 137 in order to obtain information or results related to model configuration and/or model service selection. Further, according to this embodiment, database(s) external to database management system and model server system 137 may provide information or capabilities required for model configuration and/or model service selection to model advisor 149 or other component(s) of model server system 137.

FIG. 1B illustrates a schematic diagram 180 of learning models relative to a session proxy, according to an embodiment. As shown, client 185 ₁ To 185 of _n May be via the application interface 121 ₁ To 121 _n And model instance 123 ₁ To 123 to _n Interaction, application interface 121 ₁ To 121 _n Is presented at the respective client computing system via the respective client application. Model advisor 149 may pass model instance 123 by receiving data from model system interface 135 and transmitting data to model system interface 135 ₁ To 123 to _n With client 185 ₁ To 185 of _n And (5) interaction. Accordingly, it isModel system interface 135 may facilitate clients 185 ₁ To 185 of _n Communication with model advisor 149. Further, model advisor 149 may obtain external data from external sources 165 via model system interface 135 or may otherwise communicate with external sources 165. In particular, model advisor 149 may obtain such external data via plug-in 195, which may facilitate data extraction from external sources 165. For example, one or more plug-ins 195 may facilitate the extraction of user content (e.g., user comments, photos, etc.) from a social network or social application so that the content may be transmitted to the model advisor 149 for analysis and/or processing. In addition, model advisor 149 can communicate with external model service 167 via model system interface 135. According to alternative embodiments, the model advisor 149 may obtain external data directly from external sources 165 and/or external model services 167.

Furthermore, as shown in fig. 1B, model system interface 135 may facilitate client 185 ₁ To 185 of _n Communication with a repository collection 157. Further, model advisor 149 may be in communication with one or more of the repository sets 157. In particular, reservoir 157 ₁ May be included with client 185 ₁ To 185 of _n A client data store of one or more associated data (such as past client inputs, comments, etc.). Reservoir 157 ₂ May be a knowledge base repository including knowledge base(s). These knowledge base(s) may include records of various solutions corresponding to respective problems (e.g., may be via model instance 123 ₁ To 123 to _n Client queries, questions, and/or tasks received from one or more client computing systems). These knowledge base(s) further may include session proxy learning model based and clients 185 ₁ To 185 of _n Analysis information for one or more aspects of interactions between one or more of the(s), and analysis information related to general aspects of the session agent learning model, including repository data, external data, and client information (e.g., a server analysis). In addition, model advisor 149 may request in response to client servicesThese knowledge base(s) are consulted during the process of selecting at least one model service. Reservoir 157 ₃ May be an event/modification repository that may include event and modified database(s). The database(s) of events and modifications may include all outstanding issues and/or records of resolved issues, including records relating to any service ticket that was previously or currently opened to resolve the corresponding client service request. Reservoir 157 ₄ May be a log repository including one or more logs. The one or more logs may each include one or more clients 185 ₁ To 185 of _n A record of the tracked information exchange with the model server system 137. Store 157 in store collection 157 ₁ To 157 of ₄ Merely as an example. In alternative embodiments, a repository set 157 may include a repository subset 157 ₁ To 157 of ₄ And/or additional repositories.

Further, as shown in FIG. 1B, model advisor 149 may review and/or may update model policies 155. Model policies 155 may include a set of rules that indicate how and/or when model instance 123 is updated ₁ To 123 to _n Or other aspects of the session proxy learning model (such as content within repository collection 157). In an embodiment, model policies 155 may include one or more rules that specify one or more periodic update policies regarding one or more model aspects. The one or more periodic update policies may direct the updating of one or more model aspects at one or more specified periodic time intervals. Additionally or alternatively, the one or more periodic update policies may direct the updating of one or more model aspects based on changes to the knowledge base(s) and/or based on analysis information (e.g., server analysis). In yet another embodiment, the model policies 155 may include one or more rules that specify one or more management update policies. The one or more management update policies may direct the updating of one or more model aspects at the discretion of a domain administrator associated with the session proxy learning model.

In addition, model advisor 149 may obtain information from model prediction database 159 and/or may facilitate updating model prediction database 159. In particular, model advisor 149 may obtain information regarding model service(s) associated with respective stored service requests from model prediction database 159. These stored service requests may be based on previous client service requests. Model advisor 149 can facilitate providing model service(s) associated with respective stored service requests directly via model server system 137 and/or externally (e.g., via external service provider 167). In addition, model advisor 149 can facilitate the exchange of information regarding model service(s) referenced in model prediction database 159. Based on information in model prediction database 159, model advisor 149 may respond to corresponding client service requests to clients 185 ₁ To 185 of _n Provides model service prediction(s).

Model advisor 149, model system interface 135, model policies 155, repository set 157, model prediction database 159, and model instance 123 ₁ To 123 to _n Key components of a session proxy learning model (e.g., a robot model, a chat robot model, or a session model) according to various embodiments described herein may be constructed. Schematic 180 as shown in fig. 1B depicts an exemplary configuration of model components. Alternative configurations of the model components are intended to fall within the scope of the various embodiments.

Fig. 2 illustrates a method 200 for resolving client service requests via a session proxy learning model. One or more steps associated with method 200 and other methods described herein may be performed in a client-server computing environment (e.g., computing infrastructure 100). Additionally or alternatively, one or more steps associated with method 200 and other methods described herein may be performed within one or more workloads of a cloud computing environment. Additionally or alternatively, one or more steps associated with method 200 and other methods described herein may be performed in a peer-to-peer network environment, in which case one or more method steps described herein may be performed via a peer application of a peer-to-peer computing system.

A model advisor of a server system (e.g., model advisor 149 of model server system 137) may facilitate processing in accordance with method 200. Method 200 may begin at step 205, where a model advisor may receive client service requests associated with a client system (e.g., client computing system 105) and data associated with the client service requests at step 205. The client service request may be a query submitted by a client to a model system interface (e.g., model system interface 135) via a client application interface of the client system (e.g., application interface 121 of client computing system 105). In the context of the various embodiments described herein, a client may be a person, a group of individuals, or a computer-based entity that has physical and/or electronic access to a client system. Alternatively, the client service request may be submitted via a model instance (e.g., model instance 123) associated with the session proxy learning model and accessible via a client application interface of the client system. The model system interface may be a console manager or a service manager. The model system interface, along with the model instance, may form a corresponding portion of the front end of the session proxy learning model. Upon receiving the client service request, the model system interface may transmit the client service request and associated data to the model advisor, which may receive the client service request and associated data according to step 205. The data associated with the client service request may include at least one keyword extracted from the client service request and the context information.

In an embodiment, the context information may include data associated with the client system. The client system data may include client profile data and/or client asset data. The client profile data may be a data from a client data store (e.g., store 157 ₁ ) The acquisition and may relate to computing details (e.g., associated software) and/or personal details (related to a client using the client system) of the client system. Client asset data may beTo include physical characteristics of the client system including network access capabilities of the client system, network locations of the client system, and/or processing capabilities of the client system. Further, the context information may include information about at least one service request related to the client service request. The at least one related service request may have one or more similarities to the client service request, e.g., similar keyword(s) and/or similar service question (s)/topic(s). The at least one related service request may originate from the client system and/or other system(s). The model system interface may maintain a record of incoming service requests, for example, in the form of service tickets. Accordingly, the model system interface may identify a corresponding record of the service request that pertains to the client service request by accessing a service ticket or other record associated with the relevant service request. The model system interface may forward this information to the model advisor, which may in turn receive the information according to step 205. Further, model advisors may be able to access event/modification libraries (e.g., library 157 ₃ ) Information is obtained about at least one service request related to a client service request, including information about any service ticket that was previously or currently opened. The model advisor may use information regarding at least one service request related to a client service request to determine the nature and/or extent of similarity between the client service request and other service requests.

At step 210, based on the data associated with the client service request, the model advisor may select at least one model service from a plurality of model services associated with the session proxy learning model. The at least one model service selected by the model advisor may be provided directly by the server system (e.g., a "preset" model service) and/or may be provided externally in software as a service by one or more external service providers (e.g., external service provider 165). The model advisor may analyze the at least one keyword extracted from the client service request and the client profile data and/or the client asset data to select at least one model service in accordance with step 210. As discussed further herein, the model advisor may select at least one model service associated with a service request stored in a prediction database (e.g., model prediction database 159) in response to determining a correspondence between a client service request and a stored service request. In response to determining that such correspondence does not exist, the model advisor can select at least one model service by evaluating the client service request based on a plurality of predefined factors. An embodiment relating to selecting at least one model service in response to a client service request according to step 210 is described with respect to fig. 3.

While the model advisor may select a single model service in response to determining that the single model service most closely addresses the client service request, the model advisor may also select multiple model services in response to determining that the single model service cannot explicitly address the request (i.e., in response to determining that no single model service of the multiple model services has an explicit advantage over all other model services when evaluating the respective model service). This may occur if the client service request includes multiple aspects (e.g., multiple identifiable issues presented to be resolved). For example, among a plurality of model services, the model advisor may determine that model service A best addresses one aspect of the client service request, but model service B best addresses another aspect of the client service request. In response, the model advisor can select both model service A and model service B according to step 210.

At step 215, the model advisor can provide a selection of at least one model service to the client system in response to the client service request. The model system interface may facilitate providing the selected at least one model service to the client system. In an embodiment, the model advisor may provide the selection to the client system via a model instance associated with the session proxy learning model and accessible by the client system, or via another aspect or portion of the client application interface. The selected at least one model service provided in accordance with step 215 may constitute a prediction of at least one model service that may be capable of most appropriately resolving the client service request. Based on the prediction, a client associated with the client system may attempt to solve a problem in the client service request via the provided at least one model service. In particular, the client may attempt to access at least one model service provided by the model advisor at step 215 in order to obtain a solution to the client service request (e.g., one or more solutions to the problem reflected in the client service request). The client may attempt to review the provided at least one model service via the model instance.

Fig. 3 illustrates a method 300 for selecting at least one model service among a plurality of model services associated with a session proxy learning model, according to an embodiment. Method 300 provides an exemplary embodiment with respect to step 210 of method 200. Method 300 may begin at step 305, where a model advisor may determine whether a stored service request of a plurality of stored service requests corresponds to a client service request. Multiple stored service requests may be referenced in the prediction database. According to an embodiment, the plurality of stored service requests may include one or more stored queries. An embodiment is described with respect to fig. 4 regarding determining whether a stored service request of a plurality of stored service requests corresponds to a client service request according to step 305. In response to determining at step 305 that a stored service request of the plurality of stored service requests corresponds to a client service request, at step 310, the model advisor may select at least one model service of the plurality of model services that is associated with the stored service request. In response to determining at step 305 that none of the plurality of stored service requests corresponds to a client service request, at step 315 the model advisor may select at least one model service from the plurality of model services by evaluating the client service request based on a plurality of predefined factors. An embodiment is described with respect to fig. 5 regarding the selection of at least one model service by evaluating a client service request based on a plurality of predefined factors according to step 315. In the context of this embodiment, the predefined factors may include social media trends, cost variable(s), and/or client configuration.

Fig. 4 illustrates a method 400 for determining whether a stored service request of a plurality of stored service requests corresponds to a client service request received at step 205, according to an embodiment. Method 400 provides an exemplary embodiment with respect to step 305 of method 300. Method 400 may begin at step 405, where a model advisor may determine whether a stored service request meets input criteria associated with a client service request.

In an embodiment, the input criteria analyzed according to step 405 may include keyword requirements. To determine whether the stored service request meets the keyword requirements, the model advisor may compare any keyword(s) provided for the stored service request with at least one keyword extracted from the client service request. For example, the model advisor may determine whether any keyword(s) associated with a stored service request are the same as or synonymous with at least one keyword extracted from a client service request. Additionally or alternatively, the model advisor may determine whether any keyword(s) associated with a stored service request have an ontological relationship to at least one keyword extracted from a client service request.

In yet another embodiment, the input criteria analyzed according to step 405 may include subject matter requirements. The topic requirements may be based on classification of service requests according to topic. To determine whether the stored service request meets the classification requirement, the model advisor can compare any topic classification of the stored service request with any topic classification of the client service request to determine topic compatibility. For example, the model advisor may determine whether any topic classification associated with a stored service request is the same as or synonymous with any topic classification associated with a client service request. Additionally or alternatively, the model advisor may determine whether any topic classification associated with a stored service request has an ontological relationship to any topic classification associated with a client service request. Any topic classification associated with a client service request may be specified explicitly by a client via a client system and/or may be determined by a model advisor upon receipt of a client service request.

The model advisor may accomplish the keyword comparisons and/or the topic comparisons by using text comparisons (e.g., via one or more string matching algorithms and/or text comparison algorithms) and/or natural language processing. In response to determining at step 405 that the stored service request meets the input criteria associated with the client service request, the model advisor may proceed to step 410. In response to determining at step 405 that the stored service request does not meet the input criteria associated with the client service request, the model advisor may proceed to step 415 where the model advisor may determine that the stored service request does not correspond to the client service request.

At step 410, the model advisor may determine that the stored service request meets at least one predetermined similarity threshold with respect to the client service request. The at least one predetermined threshold may ensure that aspects of the stored service request are within a similarity threshold of corresponding aspects of the client service request. In an embodiment, the model advisor may make the threshold determination by confirming that any comparison(s) made against the input criteria at step 405 indicate a minimum level of similarity between stored service requests and client service requests. Although a comparison indicating the same relationship between aspects of stored service requests and client service requests may almost always satisfy at least one predetermined similarity threshold, some comparisons indicating synonymous or ontological relationships may not satisfy at least one predetermined similarity relationship. At least one predetermined similarity relationship may be specified by a model administrator. Additionally or alternatively, the model advisor may provide the client with the option to calibrate at least one predetermined threshold. In response to determining at step 410 that the stored service request meets at least one predetermined similarity threshold with respect to the client service request, the model advisor may proceed to step 420, at step 420, the model advisor may determine that the stored service request corresponds to the client service request. In response to determining at step 410 that the stored service request does not meet at least one predetermined similarity threshold with respect to the client service request, the model advisor may proceed to step 415, at step 415, the model advisor may determine that the stored service request does not correspond to the client service request.

In an embodiment, the model advisor may process respective ones of the plurality of stored service requests according to method 400 until a stored service request corresponding to a client service request is found. In an alternative embodiment, the model advisor may process each stored service request of the plurality of stored service requests according to method 400. According to this alternative embodiment, a model advisor may determine a plurality of stored service requests corresponding to a client service request according to method 400.

Fig. 5 illustrates a method 500 of selecting at least one model service among a plurality of model services by evaluating a client service request received at step 205 based on a plurality of predefined factors, according to an embodiment. Method 500 provides an example embodiment with respect to step 315 of method 300. While social media trends, cost variable(s), and client configuration are predefined factors discussed in the context of method 500, model advisor may evaluate client service requests based on additional or alternative factors. The model advisor can evaluate client service requests by: by measuring classified data associated with a client service request based on a plurality of predefined factors and calculating a respective score for the classified data in the context of each of a plurality of model services associated with a session proxy learning model.

Method 500 may begin at step 505, where a model advisor may categorize data associated with a client service request at step 505. The model advisor may categorize the client service request data by analyzing at least one keyword extracted from the client service request and/or the contextual information. Additionally or alternatively, the model advisor may categorize the client service request data by determining and categorizing at least one other aspect of the client service request. In particular, the model advisor can categorize client service request data by identifying: at least one topic associated with the client service request, at least one intent associated with the client service request, at least one entity associated with the client service request, and/or at least one potential action (e.g., solution) for resolving the at least one intent. In the context of the various embodiments described herein, intent may indicate a purpose or goal. An entity in the context of various embodiments may be an object class or data type that enables at least one (potential) action to be selected in order to address one or more intents. After classifying the data associated with the client service request according to step 505, the model advisor may store the classified data in the data associated with the model advisor (e.g., model advisor data 153). Additionally or alternatively, the model advisor may store the categorized data in a client data store.

For example, assuming that the model advisor receives a client service request in the form of a bill payment request from a client tax payer, the model advisor may identify financial topics and bill payment intents. The model advisor may further identify the bill as an entity. Identifying the bill as an entity may in turn enable the model advisor to identify potential actions on the bill in order to resolve bill payment intent, e.g., to process payment of the bill on behalf of the client tax payer. After classifying the data associated with the bill payment request from the client tax payer according to step 505, the model advisor may store the identified topics, intents, entities, and potential actions in the data associated with the model advisor and/or in the client data store.

At step 510, the model advisor may measure categorized data associated with client service requests based on social media trends. The model advisor may observe social media trends by monitoring social channels (e.g., through a collection of social media providers). The set of social media providers may include a specified set of social network applications. Observing trends within a collection of social media providers may enable model advisors to measure categorized data within a control environment, thus providing a basis for equal comparison of multiple model services. In particular, the model advisor may calculate social media scores for the categorized data within the set of social media providers in the context of each of the plurality of model services. The social media score may be based on one or more scores calculated with respect to social media moods, social media popularity, and social media scope. An embodiment is described with respect to measuring categorized data (including the calculation of social media scores) based on social media trends with respect to FIG. 6.

Optionally, at step 515, the model advisor may measure classified data associated with the client service request based on the at least one cost variable. The model advisor can compare respective costs of each of the plurality of model services in terms of resolving or otherwise processing at least one aspect of the categorized data (e.g., at least one topic, at least one intent, at least one entity, at least one potential action, and/or at least one keyword). For example, based on the classified data, the model advisor may compare the resource utilization costs with respect to the identification and/or use of each of the plurality of model services. In particular, identifying and analyzing one or more of the plurality of model services may incur costs relative to one or more search services. Based on the measurements at step 515, the model advisor can calculate cost variable scores for the classified data in the context of each of the plurality of model services.

Optionally, at step 520, the model advisor may measure classified data associated with the client service request based on the client configuration. The model advisor may analyze at least one aspect (e.g., at least one topic, at least one intent, at least one entity, at least one potential action, and/or at least one keyword) of the classified data in the context of the network location and/or the asset of the client system to determine which of the plurality of model services can most properly address the at least one aspect. For example, a particular model service may most appropriately address a certain topic or intent identified within the categorized data, taking into account the location and/or assets of the client system. Based on the measurements at step 520, the model advisor can calculate client configuration scores for the categorized data in the context of each of the plurality of model services.

At step 525, the model advisor may select at least one model service from the plurality of model services in response to measuring the classified data associated with the client service request. The selection of at least one model service according to step 525 may constitute a prediction of the model service(s) best suited to address the client service request received at step 205. In an embodiment, the model advisor may select at least one model service among the plurality of model services by comparing model service scores calculated for the classified data, respectively, in the context of each of the plurality of model services in response to measuring the classified data. According to this embodiment, the model advisor may calculate model service scores for the categorized data in the context of each of the plurality of model services, and may then compare the respective calculated model service scores. For example, the model advisor may iteratively compare each model service score to each other model service score. Further, according to this embodiment, the model advisor may select at least one model service in the context of which the model advisor calculates the highest model service score for the classified data. In particular, the model advisor may select a single model service in response to determining that the model service score calculated for the classified data is highest in the context of the single model service. The model advisor may select the plurality of model services in response to determining that the model service score calculated for the classified data is highest in the context of the plurality of model services.

The model service score calculated for the categorized data in the context of each of the plurality of model services may be based on the social media score calculated according to step 510. Further, the model service score calculated for the classified data in the context of each of the plurality of model services may optionally be based on the cost variable score calculated according to step 515. Further, the model service score calculated for the classified data in the context of each of the plurality of model services may optionally be based on the client configuration score calculated according to step 520. If the model service score calculated for the categorized data in the context of each of the plurality of model services is based on more than one of the social media score calculated according to step 510, the cost variable score calculated according to step 515, and the client configuration score calculated according to step 520, the model advisor may calculate the model service score by averaging the respective scores or summing the respective scores. In an embodiment, the model advisor may calculate the model service score by: the social media scores, cost variable scores, and client configuration scores are weighted evenly and then averaged or aggregated (or a subset of the three scores). In an alternative embodiment, the model advisor may calculate the model service score by: different weighting values are applied to the social media scores, the cost variable scores, and the client configuration scores and then averaged or aggregated (or a subset of the three scores averaged or aggregated) based on the non-uniform weighting. By applying non-uniform weighting according to this alternative embodiment, the model advisor may emphasize one or more predefined factors relative to other predefined factor(s). For example, upon determining that social media trends are relatively more important than cost variable(s) or client configuration relative to the categorized data, the model advisor may calculate model service scores by applying higher weights to social media scores relative to cost variable scores and client configuration scores.

After selecting at least one model service according to step 525, the model advisor may store associations between client service requests and the selected at least one model service in a predictive database at step 530. By storing the association in a prediction database, the model advisor may provide predictions including the selected at least one model service upon receipt of a subsequent client service request corresponding to the client service request, thus avoiding unnecessary repetition of steps 505-525. Further, at step 530, the model advisor may store the client service request (or a reference thereto) in the prediction database along with an association between the client service request and the selected at least one model service. For example, in the context of method 300, in a subsequent session, the model advisor may determine at step 305 that the subsequent client service request corresponds to the client service request as stored in accordance with step 530, and at step 310, the model advisor may select at least one model service associated with the stored client service request after storing the association in accordance with step 530.

FIG. 6 illustrates a method 600 of measuring categorized data associated with a client service request based on social media trends, according to an embodiment. Method 600 provides an example embodiment with respect to step 510 of method 500. Method 600 may begin at step 605, where a model advisor may calculate emotion scores for at least one aspect of classified data within a set of social media providers in the context of each of a plurality of model services. Emotion associated with aspects within a collection of social media providers in the context of the various embodiments described herein may refer to feeling, emotion, and/or attitude. In particular, the model advisor may calculate the emotion score based on the emotion levels (e.g., high/low emotion) and emotion polarities (e.g., positive/neutral/negative emotion) associated with at least one aspect (e.g., at least one topic, at least one intent, at least one entity, at least one potential action, and/or at least one keyword) of the categorized data in the context of each of the plurality of model services. For example, for each of a plurality of model services, the model advisor may calculate emotion scores having an emotion level portion and an emotion polarity portion to evaluate any reference(s) to at least one aspect of the categorized data within the set of social media providers in the context of the model service. In this example, an emotional score indicating a highly active social media emotion relative to at least one aspect in the context of a certain model service may indicate a relatively higher likelihood that the model service may properly address at least one aspect of the categorized data associated with the client service request. Conversely, in this example, an emotion score that indicates a highly negative social media emotion relative to at least one aspect in the context of a model service may indicate a relatively low likelihood that the model service may properly address the at least one aspect. In an alternative embodiment, for each of a plurality of model services, the model advisor may calculate an emotion score having an emotion level portion or an emotion polarity portion (but not both) to evaluate any reference(s) to at least one aspect of the categorized data within the social media provider collection in the context of the model service.

In yet another embodiment, the model advisor can calculate a plurality of respective mood scores for a plurality of aspects of the categorized data within the set of social media providers in the context of each of a plurality of model services. According to this embodiment, the model advisor may average or aggregate multiple respective emotion scores for the categorized data in the context of each of the multiple model services. For example, if the model advisor calculates one emotional score relative to topic A and another emotional score relative to topic B in the context of each of the plurality of model services, the model advisor may average the emotional scores calculated relative to topic A and relative to topic B for the categorized data in the context of each of the plurality of model services or aggregate the emotional scores into a single emotional score.

At step 610, the model advisor may calculate popularity scores for at least one aspect of the categorized data within the set of social media providers in the context of each of the plurality of model services. The popularity of an aspect within a set of social media providers in the context of the various embodiments described herein may refer to an advantage or significance of that aspect within the set of social media providers. In particular, the model advisor may calculate popularity scores within the context of each of a plurality of model services based on a plurality of references to at least one aspect of the categorized data (e.g., at least one topic, at least one intent, at least one entity, at least one potential action, and/or at least one keyword) within the set of social media providers. For example, for each of a plurality of model services, the model advisor may calculate a popularity score reflecting popularity (e.g., a ratio of a number of references to at least one aspect in content posts within a social media provider set to a total number of content posts within the social media provider set in the context of the model service). In this example, a higher popularity (and thus a higher popularity score) calculated in the context of a certain model service relative to at least one aspect of the classified data may indicate a relatively higher likelihood that the model service may properly address at least one aspect of the classified data associated with the client service request. Conversely, in this example, a lower popularity (and thus a lower popularity score) calculated in the context of a certain model service relative to at least one aspect of the classified data may indicate a relatively lower likelihood that the model service may properly address the at least one aspect.

In yet another embodiment, the model advisor can calculate a plurality of respective popularity scores for aspects of the categorized data within the set of social media providers in the context of each of a plurality of model services. According to this embodiment, the model advisor may average or aggregate multiple respective popularity scores into a single popularity score for the categorized data. For example, if the model advisor calculates one popularity score with respect to topic C and another popularity score with respect to topic D in the context of each of the plurality of model services, the model advisor may average the popularity scores calculated with respect to topic C and with respect to topic D for the categorized data in the context of each of the plurality of model services or aggregate the popularity scores into a single popularity score.

At step 615, the model advisor may calculate a scope score for at least one aspect of the categorized data within the set of social media providers in the context of each of the plurality of model services. In the context of the various embodiments described herein, the scope of an aspect within a set of social media providers may refer to the scope or magnitude of the impact of that aspect within the set of social media providers. In particular, the model advisor may calculate the range score within the context of each of the plurality of model services based on a plurality of users discussing at least one aspect of the categorized data (e.g., at least one topic, at least one intent, at least one entity, at least one potential action, and/or at least one keyword) within the set of social media providers. For example, for each of a plurality of model services, the model advisor may calculate a scope score reflecting a scope ratio (e.g., a ratio of a number of users discussing at least one aspect within a set of social media providers to a total number of users within the set of social media providers in the context of the model service). In this example, a higher range ratio (and thus a higher range score) calculated relative to at least one aspect in the context of a certain model service may indicate a relatively higher likelihood that the model service may properly address at least one aspect of the categorized data associated with the client service request. Conversely, in this example, a lower range ratio (and thus a lower range score) calculated relative to at least one aspect in the context of a certain model service may indicate a relatively lower likelihood that the model service may properly address the at least one aspect.

In yet another embodiment, the model advisor can calculate a plurality of respective scope scores for aspects of the categorized data within the set of social media providers in the context of each of a plurality of model services. According to this embodiment, the model advisor may average or aggregate multiple respective range scores into a single range score for the classified data. For example, if the model advisor calculates one range score relative to topic E and another range score relative to topic F in the context of each of the plurality of model services, the model advisor may average the calculated range scores relative to topic E and relative to topic F for the categorized data in the context of each of the plurality of model services or aggregate the range scores into a single range score.

In an embodiment, the model advisor may calculate the mood score, popularity score, and/or range score in the context of each of the plurality of model services by focusing on the same aspect(s) of the classified data for the purpose of each calculation. For example, a model advisor may calculate emotion scores, popularity scores, and scope scores in the context of each of a plurality of model services by focusing on topic a within the categorized data. In alternative embodiments, the model advisor may calculate the mood score, popularity score, and/or range score in the context of each of the plurality of model services by focusing on different aspects of the categorized data for the purpose of each calculation. For example, a model advisor may calculate emotion scores, popularity scores, and scope scores in the context of each of a plurality of model services by: the intent C and the entity D within the classified data are concentrated to calculate the mood score, the topic B within the classified data is concentrated to calculate the popularity score, and the intent C and the entity D within the classified data are concentrated to calculate the range score.

At step 620, in the context of each of the plurality of model services, the model advisor may calculate social media scores for the categorized data based on all or a subset of the respective scores (i.e., the mood score, popularity score, and/or scope score) calculated at steps 605-615. If the social media score calculated in the context of each of the plurality of model services is based on more than one of the mood score calculated according to step 605, the popularity score calculated according to step 610, and the range score calculated according to step 615, the model advisor may calculate the social media score by averaging or aggregating the respective scores. In an embodiment, the model advisor may calculate social media scores by: the mood score, popularity score, and range score are weighted evenly and then averaged or aggregated (or a subset of the three scores is averaged or aggregated). In an alternative embodiment, the model advisor may calculate the social media score by: different weighting values are applied to the emotion score, popularity score, and scope score, respectively, and then the three scores are averaged or aggregated (or a subset of the three scores is averaged or aggregated) based on the non-uniform weighting. By applying the non-uniform weighting according to this alternative embodiment, the model advisor may emphasize one or more social media trends relative to other social media trend(s). For example, upon determining that social media moods are relatively more important than social media popularity or social media scope relative to the categorized data, the model advisor may calculate the social media score by applying a higher weight to the mood score relative to the popularity score and scope score.

In yet another alternative embodiment, the social media score may include a score(s) calculated relative to one or more additional social media trends. The social media score calculated for a given model service may indicate a likelihood that the model service may properly address at least one aspect of the categorized data associated with the client service request. The model advisor may perform all or a subset of the steps 605-615 of method 600. Additionally, the model advisor may perform steps 605-615 in any order. By computing one or more scores in order to measure social media trends according to method 600, a model advisor can determine relative compatibility of one or more aspects associated with a client service request in the context of each of a plurality of model services.

FIG. 7 illustrates a schematic diagram 700 of calculating model service scores 705 for classified data relative to the context of each of a plurality of model services, according to an embodiment. The model advisor may calculate model service scores 705 for the categorized data in the context of each of the plurality of model services according to the previously described embodiments of step 525. In particular, the model advisor may calculate model service scores 705 based on social media scores 710 calculated according to step 510. Optionally, the model advisor may further calculate the model service score 705 based on the cost variable score 715 calculated according to step 515. Optionally, the model advisor may further calculate the model service score 705 based on the client configuration score 720 calculated according to step 520. If the model service score 705 is based on more than one of the social media score 710, the cost variable score 715, and the client configuration score 720, the model advisor may calculate the model service score 705 by averaging out the respective scores or aggregating the respective scores. As previously described, the respective scores may be weighted uniformly or non-uniformly.

According to embodiments previously described with respect to method 600, a model advisor may calculate social media scores 710 for categorized data in the context of each of a plurality of model services according to step 620 based on the following scores: a mood score 725 calculated according to step 605, a popularity score 730 calculated according to step 610, and/or a range score 735 calculated according to step 615. If social media score 710 is based on more than one of mood score 725, popularity score 730, and range score 735, then the model service may calculate social media score 710 by averaging the respective scores or aggregating the respective scores. As previously described, the respective scores may be weighted uniformly or non-uniformly.

The description of the various embodiments of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the disclosed embodiments. All modifications and equivalent arrangements of the described embodiments should be considered to be within the scope of the invention. The scope of the invention should, therefore, be construed broadly in accordance with the appended claims, in conjunction with the detailed description, and should be construed to cover all possible equivalent variations and equivalent arrangements. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to best explain the principles of various embodiments, the practical application, or the improvement of technology over the technology found in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments described herein.

Claims

1. A method, comprising:

receiving a client service request associated with a client system and data associated with the client service request, wherein the data associated with the client service request includes at least one keyword extracted from the client service request and context information;

selecting at least one model service from a plurality of model services associated with a session proxy learning model based on the data associated with the client service request, wherein selecting at least one model service comprises: in response to determining that none of the plurality of stored service requests referenced in the predictive database corresponds to the client service request, selecting at least one model service from the plurality of model services by evaluating the client service request by:

classifying the data associated with the client service request;

measuring categorized data based on a plurality of predefined factors, the plurality of predefined factors including social media trends, at least one cost variable, and/or client configuration;

selecting at least one model service from the plurality of model services in response to measuring the classified data; and

Storing an association between the client service request and the selected at least one model service in the predictive database; and

providing the selected at least one model service to the client system in response to the client service request.

2. The method of claim 1, wherein the context information comprises data associated with the client system and information regarding at least one service request related to the client service request.

3. The method of claim 1, wherein selecting at least one model service further comprises:

in response to determining that a stored service request of a plurality of stored service requests referenced in a predictive database corresponds to the client service request, at least one model service associated with the stored service request is selected from the plurality of model services.

4. The method of claim 3, wherein determining that a stored service request of the plurality of stored service requests corresponds to the client service request comprises:

determining that the stored service request meets input criteria associated with the client service request; and

Determining that the stored service request meets at least one predetermined similarity threshold with respect to the client service request.

5. The method of claim 1, wherein selecting at least one model service among the plurality of model services in response to measuring the classified data comprises: comparing model service scores calculated for the classified data, respectively, in the context of each of the plurality of model services.

6. The method of claim 1, wherein measuring the categorized data based on the social media trends comprises:

an emotional score for at least one aspect of the classified data is calculated within a set of social media providers in the context of each model service of the plurality of model services.

7. The method of claim 1, wherein measuring the categorized data based on the social media trends comprises:

a popularity score for at least one aspect of the categorized data is calculated within a set of social media providers in the context of each of the plurality of model services.

8. The method of claim 1, wherein measuring the categorized data based on the social media trends comprises:

A scope score for at least one aspect of the categorized data is calculated within the set of social media providers in the context of each of the plurality of model services.

9. A computer readable storage medium storing a computer program having program instructions embodied therewith, the program instructions being executable by a computing device to cause the computing device to perform method steps of the method according to any one of claims 1 to 8.

10. A system, comprising:

a processor; and

memory storing an application program which, when executed on the processor, performs the method steps of the method according to any one of claims 1 to 8.

11. An apparatus comprising modules individually configured to perform each step of the method of any one of claims 1 to 8.